M I C H A E L R . G A L B R E T H , P H . D .
U S C / S Y R A C U S E S U P P L Y C H A I N I M P R O V E M E N T C O N F E R E N C E , O C T . 2 0 1 4
USING
ANALYTICS
TO IMPROVE
HIPPO
“If we have data, let’s look at data. If all we have are
opinions, let’s go with mine.”
EMERGING DECISION APPROACH:
DATA-DRIVEN
•
Why now? Data has become extremely inexpensive
to capture and store
•
Exciting shift in business: less decision making based
on HIPPO, more data-driven (LinkedIn, Google,
etc.)
DO YOU HAVE THE RIGHT DATA?
COLLECTING FIELD DATA
•
Analysis of bullet holes in WWII planes after an
enemy encounter
• Analyze the damage – this can help us determine where to
put additional armor.
Final decision – holes indicate areas where armor is
DO YOU HAVE THE RIGHT DATA?
SURVEY DATA
•
Ask lots of questions about how the
data was collected!
•
People who read email differ from
people who do not read email.
•
People with land lines differ
DO YOU HAVE THE RIGHT DATA?
IS IT CLEAN?
Data exploration before modeling begins is
critical:
Why were some customers active for 31 days in February,
but none were active for more than 28 days in January?
System bug – month codes were swapped
Why were so many customers born in 1911? Are they
really that old?
Birth date required, and holding down the “1” key is
easy for data entry personnel
Why are there negative numbers in the sales price field?
DATA ANALYTICS: A DOUBLE-EDGED SWORD?
THREE BROAD CATEGORIES OF
ANALYTICS
1. Descriptive Analytics – what has
happened
•
gathering, summarizing, and visualizing
large datasets to reveal new business
insights. In SCM:
•
Real-time data sources – GPS, RFID
•
Supply chain mapping (free.sourcemap.com)
THREE BROAD CATEGORIES OF
ANALYTICS
2. Predictive Analytics – what is likely to
happen
•
using historical data to predict future
outcomes; can use either statistical or
machine learning (computer science)
methods. In SCM:
•
Demand forecasting at all echelons
•
Emerging approach: combine multiple
THREE BROAD CATEGORIES OF
ANALYTICS
3. Prescriptive Analytics – what should
happen
•
using operations research tools to
determine which decisions will lead to the
best outcomes. In SCM:
DESCRIPTIVE ANALYTICS IN SCM:
SENSORS AND RFID
•
Manufacturing stores more data than any
other sector – sensors on plant equipment or
sold products
•
Sensors enable a product to “phone home” its
usage patterns, etc. Manage defects, service
needs, NPD, upselling
•
Projected growth in RFID tags from 2011 to
2021:
•
From 12 million to 209 billion (~17,400x)
•
“Performance transparency” – real-time
monitoring of performance at a
granular level
DESCRIPTIVE ANALYTICS IN SCM:
GPS
•
US DOT – in the US, trucking makes up 80% of total
transportation costs, delivering 54 million tons per day,
representing tens of billions of dollars.
•
Rapid adoption of
GPS
vehicle tracking devices
(currently 6 million in use):
• Usage monitoring – eliminate personal use and improve driving
(“like having a manager in every vehicle” -LiveView GPS). Incentivize desired driver behavior.
• Delivery monitoring, which can also identify profitable
backhaul opportunities in real time
• Recovery of stolen assets
PREDICTIVE ANALYTICS IN SCM:
SUPERVISED PREDICTION
Regression
Decision
Trees
Neural
Networks
Time Series
THE POWER OF PREDICTIVE MODELING
•
Early application of big data to predictive
modeling: consumer behavior
•
Predicting missing a credit card payment
• Very likely to miss: anyone who buys…
• cheap, generic motor oil • chrome skull car accessories
• Very UNLIKELY to miss: anyone who buys…
• Felt pads for chair legs
• Carbon monoxide monitors • Premium birdseed
•
Kroger is realizing revenue of over $100 million per
year selling its shopper data to consumer goods
companies
PREDICTIVE ANALYTICS IN SCM
•
Demand forecasting
•
Google Trends
•
Emerging approach – leverage social
PRESCRIPTIVE ANALYTICS IN SCM
•
Optimization
• Staffing
• Real-time vehicle routing (further leverages GPS)
• billions $ of time and fuel are wasted each year sitting in traffic
• Waste Management vehicle routing - $44M annual savings
•
Simulation
• R&D – leverage data from multiple systems (CAD, production,
etc.) to drive simulation models. Toyota now claims to have eliminated 80% of product defects before the first physical prototype is built.
• Manufacturing – create a “digital factory” (McKinsey), a very
detailed model of the entire production facility – machines, people, equipment – that enables extensive what-if analysis in a virtual environment (automotive, steel, semiconductor,
industries)
•
Emerging Idea: Open innovation
• crowdsource new ideas – from logos to product design
